Fish-eye Lens Human Detection via Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting humans in images captured by fish-eye lenses are slow and inaccurate due to the need to analyze multiple dictionaries indicating human characteristics.
Innovation Solution
A detection device that divides the image into multiple areas, estimates the probability of each area containing a human, and determines the presence of a human using a trained model, eliminating the need for multiple dictionaries by verifying the correctness of human detection through area classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple dictionaries indicating human characteristics are used for detection, then detection accuracy is improved, but detection time is lengthened
Solution Approach 1:
The patent divides the captured image into multiple areas and performs detection processing on each area separately using a single dictionary. This segmentation approach allows parallel processing of different image regions, reducing the overall detection time while maintaining the use of multiple human characteristics for accurate detection in each region.
Solution Approach 2:
The patent applies detection processing to only specific areas of the image where human detection is expected, rather than uniformly processing the entire image with multiple dictionaries. This partial action approach reduces computational load and detection time while focusing resources on relevant regions.
2Loss of time
If a single dictionary indicating human characteristics is used for detection, then detection time is reduced, but detection accuracy deteriorates
Solution Approach 1:
The patent segments the image into multiple areas and applies detection processing to each area. By combining the results from multiple area-based detections using a single dictionary, the system achieves both fast processing and improved accuracy through aggregated detection results across different regions.
Solution Approach 2:
The patent introduces a spatial dimension by dividing the image into multiple areas and performing detection in each area separately. This dimensional approach allows the system to maintain fast single-dictionary processing while achieving multi-dictionary-level accuracy through the combination of area-based detection results.
3Measurement precision
If multiple dictionaries indicating human characteristics are used for detection, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent simplifies the system by eliminating the need for multiple dictionaries while maintaining multiple area-based detection processing. This segmentation approach reduces dictionary management complexity while preserving detection accuracy through spatial distribution of detection tasks.
Solution Approach 2:
The patent makes a single dictionary serve multiple functions by applying it to multiple image areas. This universal approach reduces the need for multiple specialized dictionaries, simplifying the system while maintaining comprehensive detection capabilities across different regions.
Data Source
AI summary
A detection device detects an imaged human from a captured image captured by a ceiling camera using a fish-eye lens, and includes a division part, an area estimation part, and a determination part. The division part divides the captured image into a plurality of areas. The area estimation part calculates a probability that each of the plurality of areas includes a position where a detection target estimated to be a human is present. The determination part determines whether the detection target is a human using the probability of each of the plurality of areas.